en
Scientific article
Open access
English

Robust Inference for Generalized Linear Models

Published inJournal of the American Statistical Association, vol. 96, no. 455, p. 1022-1030
Publication date2001
Abstract

By starting from a natural class of robust estimators for generalized linear models based on the notion of quasi-likelihood, we define robust deviances that can be used for stepwise model selection as in the classical framework. We derive the asymptotic distribution of tests based on robust deviances, and we investigate the stability of their asymptotic level under contamination. The binomial and Poisson models are treated in detail. Two applications to real data and a sensitivity analysis show that the inference obtained by means of the new techniques is more reliable than that obtained by classical estimation and testing procedures.

Keywords
  • Binomial regression
  • Influence function
  • M-estimators
  • Model selection
  • Poisson regression
  • Quasi-likehood
  • Robust deviance
  • Robustness of efficiency
  • Robustness of validity
Citation (ISO format)
CANTONI, Eva, RONCHETTI, Elvezio. Robust Inference for Generalized Linear Models. In: Journal of the American Statistical Association, 2001, vol. 96, n° 455, p. 1022–1030. doi: 10.1198/016214501753209004
Main files (1)
Article (Accepted version)
accessLevelPublic
Identifiers
ISSN of the journal0162-1459
807views
1201downloads

Technical informations

Creation09/07/2012 10:50:00 AM
First validation09/07/2012 10:50:00 AM
Update time03/14/2023 5:40:48 PM
Status update03/14/2023 5:40:47 PM
Last indexation05/02/2024 12:39:02 PM
All rights reserved by Archive ouverte UNIGE and the University of GenevaunigeBlack